Slightly edited March 2014.

Content of this page composed 2004.......

Fig 1. Picture of  two 100W UL-AB1 mono amps.
General description.

Fig 2.  Sheet 1 power amp schematic, 2004, hand drawn.
Fig 3.  Sheet 2 power supply schematic, 2004, hand drawn.

Fig 4.  Sheet 3 bias and protection circuits, 2004, hand drawn.
Fig 5.
  Power vs load graphs.
Fig 6.  THD graphs
Full explanations of all included.

Please excuse my hand drawn schematics. I was over worked and underpaid during
most of my audiological career and I didn't have time in early 2000's to spend a day
with MS Paint to prepare a good looking schematic. BUT, I have now found a little time
to draw a number of schematics properly and include more useful info In March 2014.........

For those thinking of building a good 100W PP amp, they may skadoodle off to a 2014 page at

Fig 1.

Two 100w mono blocs on table

I sold these two amplifiers in early 2004 to a delighted customer living near
Cairns in Queensland
where there is a tropical climate.
He wished to power Quad ESL63 speakers used mainly during the
dry season because high humidity can be a problem with ESL. But the amps would power any other
type of speaker.

Chassis is aluminium, 470 mm long x 260 mm wide x 60mm high with above chassis content making
total height 220 high, including timber strip feet under each end of bottom plate to allow air flow through
chassis space and up around each tube through drilled holes around each power tube. The open
steel grilles over the tubes give just enough protection against tubes being bent over in sockets and
from falling objects.
Weight is approx 22 Kg, ( 49 Lb ), for each mono amp.
The largest transformer is the output transformer with a 75 mm stack of 44 mm tongue E&I GOSS E&I
laminations seen behind the output tubes.

The power transformer is 100 mm stack of 38 tongue mm lams, seen furthest from view. Anyone could
make this amp using the same size PT core used for OPT.

Silicon rectifiers are used, with C-L-C filtering using a choke of 2H and large value electrolytics to maintain
an impeccably low noise level in B+ rail.
The amp was sold in 2004 with matched 6CA7EH which I feel are more rugged than EL34,
and excellent sounding.

The 2004 hand drawn schematics for this amp provide ease of servicing for the amps I sold.

Fig 2.
Sheet 1, 100Watt UL AB1 power amp schematic 2004.


Fig 2 above shows the hand drawn 2004 schematic. Output tubes can be 6 x EL34 or 6CA7,
but I sold it with Russian 6 x 6AC7 which were rugged copies of US made tubes and which
produce excellent sound.

Input V1 on Fig2 is a paralleled 12AU7, but could be 12AT7, 6DJ8, 6CG7, or 6SN7, or 6SH7 in triode.
However, any change from 12AU7 for V1 may involve some careful schematic changes to V1 cathode
biasing and NVFB networks to maintain the same amount of applied dB of feedback.
V2 and V3 d
river triodes are 2 x 6SN7, with each having their two triodes paralleled, to best drive the
paralleled grid biasing resistors for EL34.

There are TWO global loops of NFB used.
One is labelled VNFB, which is just normal conventional series
global voltage negative feedback
loop where the network has 2 x 2k7 // 1nF from Vo to 50r at V1 12AU7 cathode.
The 1nF advances the phase of HF signals fed back to the V1 cathode so that there is good
control of HF amplitude above 10kHz.
The other is labelled NCFB which is a-rarely-ever-used series
global current negative feedback
loop where the speaker current passes through 0.1r shunted by 2uH inductance seen just below V1.
The CFB may seem odd, but it greatly assists the HF stability of the amp if it is ever loaded with
a pure capacitance load. The 0.1r and 2uH have pole at 8 kHz where XL = R ohms, so this network
has negligible effect at low F. This network may not be needed in an amp with a different OPT
to the one I wound. There is no Zobel R&C network used between V1 anode and 0V because the
properties of the OPT did not demand HF open loop gain be shelved to reduce the open loop
gain and phase shift.

Fixed bias is used, and there are six adjust pots and volt meter test points accessible through
6 access holes in the side of the chassis, so bias can be
adjusted easily with a cheap dc voltmeter
and screw driver without moving
the amps to a work bench or removing any covers.

V2, and V3 6SN7 driver triodes
are set up in a differential amp aka long tail pair driver/phase
inverter with their commoned cathodes connected to a "long tail" using one MJE340, which
has such a high effective collector impedance
that it can be regarded as a constant current sink.
This is easily achieved by having the emitter resistance of 47k // 4k7 unbypassed, but with base
at -25Vdc well bypassed to a -75Vdc rail. 

The CCS ensures the two oppositely phased anode signals from each 6SN7 has exactly equal
amplitude if the resistance loads of each 6SN7 are equal, which is very easy to achieve with
1% metal film resistors. The 6SN7 do not need to be matched. There is no Vac balance pot for
an owner to adjust.

The MJE340 has no sonic signature, and does no active amplification and
acts merely as a high
ac impedance sink of cathode current. It replaces
what would have been used in 1955 - a pentode
such as 6SH7 with -150Vdc supply and more complexity, but which could not perform as well
as the MJE340 as I have it.

The output of V1 input triode feeds V2 grid of the LTP via a LF gain shelving network with
1M0 // 47nF and 220k. V3 grid grounded to 0V. This network adjusts the LF open loop response
to ensure
unconditional LF stability with GNFB applied. The network produces a loss of OL gain
below 20 Hz, but lowers the ultimate LF pole to 0.27Hz. At such a low F the phase shift caused by
the CR coupling cannot cause LF oscillations because open loop gain is less than 1.0.
This LF network ensures superb LF stability, with excellent bass handling and good behaviour
after gross overload.

Zobel networks with series R&C are used across each half primary on the OPT,
1k5 + 2n2
and across the OPT secondary, 0.22uF + 15r. These networks
ensure the amp is loaded
by a resistive load at HF when typically inductive
speaker loads are used which have
increasing Z above 20kHz.

The Zobel and NFB networks ensure complete freedom from any LF or HF oscillation with any
possible load with reactive L or C loads, or with no load at all, or with a malfunctioning output tube
or with excessively low F signals at high levels which may cause OPT core saturation.
In other words, the amp is unconditionally stable, with no exceptions.

The 1N4007 seriesed silicon diodes from the OPT anode connections to 0V are to limit the
peak output voltage swing to the same voltage as B+ anode
supply voltage, ie, about 430Vpk.
Without the diodes, and without a load, there is every possibility the owner may turn up the
volume control without remembering to plug in the speaker leads. This will cause the signal
voltage generated across the OPT primary to soar to 1,500Vrms, say 3 times normal maximum
signal of 470Vrms. This puts undue pressure on OPT and wiring insulation and output tubes
and Vac arcing can occur and trigger DC arcing which may more easily maintain itself and cause
damage to OPT, tubes, and before any fuse can blow.
This phenomena of OPT voltage soaring is caused by release of energy stored in the OPT
leakage inductance; it is called a "back emf effect". I don't have time for further details but
believe me, it happens in most tube amps. These high voltage have opposite phases at
each end of primary. When each end of primary tries to become negative with respect to
0V, the diodes conduct immediately and act as a short circuit, and negative voltage swings at
each end of primary are limited to -1.8V where 3 series diodes are used. The overall effect is
that with no load, the anode signal at each end of primary becomes clipped at a peak voltage
about equal to the B+ anode supply, and the parts involved are safe from arcing damage.
At normal operation where peak V swing at anodes are always less than the B+, the diodes
have zero effect because their impedance when not conducting is a huge value.

Power output is
plotted in sheet 4, for selectable triode or UL connected amps.
Bandwidth at 72 watts, 8 ohms, is 10 Hz to 65 kHz.
Output impedance < 0.5 ohms,
Distortion < 0.3% at 100 watts, 4 ohms, and less than 0.05% at 3 watts.
Noise is very low.
The amp is fitted with the usual active protection circuits, shown on sheet 3.  
An led on the front of the chassis indicates clipping, or any fault condition.

Fig 3.
Sheet 2, 100W UL AB1 amp power supply

100w monobloc power supply schematic

The power supply has all solid state rectifiers. A CLC filter ensures low B+ rail noise
and a 50
ohm resistor limits peak charge currents into the two 220uF series input
caps to the CLC for B+.
The 50 ohms also keeps the B+ at about +410V to enable a high idle bias current
thus ensuring a large amount of class A power.
DC is applied to V1 heaters and fixed grid bias and B+ supplied to V1 is shunt regulated.
There is a relay in the power transformer HT secondary circuit which
is operated by the
protection circuit
shown in sheet 3.

Fig 4.
Sheet 3,  100W UL AB1 bias and protection schematic.
100w mono bloc bias and protect schematic.

The fixed bias adjustment pots are arranged as shown to give each tube a
range of applied
grid bias between -33V and -48V.
Each pot for each output tube grid bias s adjusted so
0.5Vdc appears across
the same tube's 10 ohm cathode resistor.
This indicates 50mA of idle current.
The process of setting the bias for the amp is repeated 3 times after the amp has
warmed up because bias adjustments are interactive.

The circuit is actively protected against bias failure in one or more tubes, or against
excessive and continued use with loads which are too low in value.

Each cathode of each output tube is grounded through 10 ohms and points K1 to K6 are
all fed to a common cathode monitoring signal path which has
its voltage reduced by the
R divider of 4k7 and 2k7 with a 220uF cap to
remove unwanted ac signals during operation.
Should one of more cathode dc currents rise to dangerous levels for more than a couple
of seconds, the SCR will trip and  the 16V relay supply will be pulled
to 0V since the
50 ohm 10W resistor is grounded at the relay end.

If the SCR is tripped the fault LED will turn on. The same LED will flash if the amp clips
since the error signal from the output of V1 is fed through
a high resistance path to a
darlington pair of bjt driving a second gain bjt to
turn on an LED.
All the protection parts will fit on a small board under the chassis and all can be easily
sourced at any electronic parts shops.

There is a 6 second delay for turn on of the B+.
Fig 5.
Sheet 4, power vs load graph.


The graphs show maximum power levels at thd <2% just before clipping
with a range
of load values.

Curves A and B are for Ultralinear with either 2k : 6 ohm load match or 4.5k : 6 ohms.
Curves C and D are for Triode with either 2K : 6 ohms or 4k5 : 6 ohms.
Any value of load along the RL bottom line can be chosen and the output power
max can be seen on the curves.

Using the higher Z ratio on the OPT results in less maximum  output
power for most
common loads between 4 and 8 ohms but the class A
portion of total power increases
and although the total AB power reduces
the THD and output impedance are both
much lower.

For example, if the OPT is set for a load match for 4k5 : 6 and an 8 ohm
speaker is
used, then the maximum UL power output only 52 watts,
but it is virtually all class A
and the THD will be about 1/2 that of 6 ohms.

Load matching is all very confusing to most people. A common
is that the more ohms a speaker has, the more difficult
it is to drive. This comes from
knowing that common sense tells us carrying 10 bricks is more difficult
than carrying
1 brick.

But in fact, more ohms mean a less difficult load. Like so many things in
electronics, basic
commonsense seems reversed to a novice. An Ohm is the unit of resistance which is a
property of a substance to resist the flow of electrons. Where you have 1Volt applied
across a length of wire, and there is 1 Amp of current, the wire is said to have resistance
= 1 ohm. Ohm's Law is expressed in a formula, R = V / I , ie, Resistance = Voltage / Current.
Power is the heat liberated in the resistance due to current flow. It is most simply calculated
as P in Watts = Volts x Amps, ie, P = V x I. To make a resistance get hot, you must generate
electrical power. The more power you want in a resistance, the harder it is to generate
that power. The lower the ohms, the higher the current flow. The higher the current flow,
the higher the power. and it applies to speakers. As speaker ohms are reduced, the current
increases and the generator must be able to provide the correct range of current and voltage.
So an amp is like a generator. It may be happy to make 16Vrms for an 8 ohm speaker to make
2Amps of current and give 32Watts of power. But if 2 ohms is used instead of 8 ohms,
the amp may not be able to make more than 3Amps, and you get only 6Vrms, and 18Watts
of power, with high THD and tubes run too hot.
If you have 16 ohms, the amp may make only 1Amp maximum, and 16Vrms, and 16Watts of
power. However, the THD will be very low, amp will run cooler, and best music is heard.
If anyone
cannot understand what Ohms are, they need to study Ohm's Law and history of
knowledge about electricity. This might lead to more questions about many other basics of
operation and load matching effects and distortions.
Books such as the
1955 4th Ed of Radiotron Designer's Handbook is an excellent read,
or they might read my other pages on tube basics. EDUCATIONAL AND DIY.

Fig 6.

Sheet 5, 100W UL AB1 THD figures for two amps.


The graphs of  total harmonic distortion, thd, are for one pair of 100 Watt tube amps
with identical schematics. There are two curves for
amps A and B and there is up to 7dB
difference between THD at below
The graphs were made during completion work on the two amps. The voltage scale is
linear and the THD scale is *logarithmic* to display
small quantities of THD more easily.
The amp A had the highest THD which is listed below with power levels
and what would be SPLs using  average modern speakers of 6 ohms and
90dB/W/M :-

1 watt , 2.45vrms, 0.02%, 90dB SPL,

2 watts, 3.46Vrms, 0.03%, 93dB SPL,
4 watts, 4.89Vrms, 0.047%, 96dB SPL,
8 watts, 6.93Vrms, 0.09%, 99dB SPL,
16 watts, 9.8Vrms, 0.15%, 102dB SPL,
32 watts, 13.8Vrms, 0.20%, 105dB SPL,
64 watts, 19.6Vrms, 0.37%, 108dB SPL,
88 watts, 23.0Vrms, 1.0% 109dB SPL, :- It is so loud, it wakes the dead.

Like nearly all amps the fidelity increases with a higher load value but with
a reduction of
maximum output power.
Most people will find that with two amplifiers average levels of 1/2 a Watt
from each will produce SPLs of 88 dB approximately although peaks in drumbeats and transients
will go a lot higher but these will be very easily
dealt with by these amps.
The average level for most loud music in a
domestic situation is 88db for men,
and 84dB for women and that includes
both channels, so indeed these amps have quite
enough power even for
teenagers who like bass boosting if possible in the preamp.
( OK, you have a killer teenager?... I don't want to know..)

Much is said about tube amp distortions spoiling performances but let's
get this into
perspective. If the speaker voltage is 5Vrms at 4 Watts of level,
then the THD = 0.05%.
Therefore the actual distortion voltage within the signal is 0.0025Vrms, and would be very
difficult to hear from across a room with speakers rated
for 87dB/W/M, because the
distortion voltage alone gives 1 micro Watt
of power, which gives an SPL that is 60dB
below the 4 watt level of 93dB,
so the distortion produces an SPL at 33 dB which would
be below the
sound level of heartbeats, breathing, and natural background sound levels.

The THD spectral voltages in tube amps such as these at low levels up to
4 Watts is usually
a mixture of predominantly 2H and 3H with other
4H, 5H, 6H, 7H, 8H, 9H etc all at least
12dB below the combined levels
of 2H and 3H indicated on the above graphs.
Either 2H or 3H harmonic
may be greater at low levels because the 2H at low levels varies
because of
unavoidable slight unbalance in the 2H currents in each half of the PP output
circuit because the 2H currents in each 1/2 of the PP circuit are slightly different in level,
thus never completely cancelling as a result of push pull action.

There are slight differences in the "matched" output tubes used in the output stage and
driver stages. Plus the input stage produces some 2H since it is a
single ended triode stage.
The 2H content is the main reason for the
differences between the two amps' THD.
But at high output levels the THD of the pair of amps becomes nearly equal and mainly 3H.
The 2H at low levels can be minimized by placing
tubes in either side of the PP circuit
so that ac signal currents measured
across the 10 ohm cathode resistors for each side
have equal totals,
or as close as one can get, and this is a very tedious thing to do and to
monitor with a distortion meter.
Unfortunately, this needs some technical expertise to achieve, and usually does not lead
to any betterment of the music.

Triode operation was tested but very little differences with THD were
recorded at the same
low levels. Changing from UL to triode operation
means the total circuit gain is reduced
about 4 db. Therefore the amount
of applied global NFB is also reduced by the same amount.
With less
global FB one would expect output resistance and distortion to rise but the triode
connection itself gives a compensating reduction in THD
and output resistance so the
change from UL to triode does not require
any change to the global NFB network
resistances and the output
impedance and THD will remain very similar in either UL or triode.

These amps were prepared for someone with QUAD ESL63 and the
amount of NFB is
not high and could have been increased to levels
used more commonly by other makers
who might use say 20dB
which would reduce all the above THD figures by about 9dB,
or to
1/3 of the figures mentioned. I felt there was no need since the output impedance
and THD was low enough.

There is nobody I know who can tell the difference between triode and UL connected
output stages if the power levels are well away
from clipping and if the general total
levels of NFB are similar as
I suggested above.

Back to Education and DIY directory.

Back to Index Page.